Is gemma-4-12B-it-qat-q4_0-unquantized or GigaChat3-10B-A1.8B-base smaller?

Both publish quantizations; compare the table for the exact sizes.

Which uses less memory at long context?

GigaChat3-10B-A1.8B-base — its KV cache at 32K context is 2.7x smaller, because of how its attention layers cache.

Model comparison

gemma-4-12B-it-qat-q4_0-unquantized vs GigaChat3-10B-A1.8B-base

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: GigaChat3-10B-A1.8B-base's KV cache at 32K is 2.7× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

	gemma-4-12B-it-qat-q4_0-unquantized	GigaChat3-10B-A1.8B-base
Parameters	12.0B	11.5B
Architecture	gemma4	deepseek2
Layers	48	26
Native context	262,144	262,144
Mixture of experts	no	yes, 64 experts
Quantizations published	2	8
Smallest quantization	6.50 GiB	6.03 GiB
Q4_K_M	—	6.03 GiB
Licence	apache-2.0	mit

KV cache by context

the term that decides long-context viability

Context	gemma-4-12B-it-qat-q4_0-unquantized	GigaChat3-10B-A1.8B-base	Ratio
4,096	0.72 GiB	0.11 GiB	6.29×
8,192	0.97 GiB	0.23 GiB	4.24×
16,384	1.47 GiB	0.46 GiB	3.21×
32,768	2.47 GiB	0.91 GiB	2.70×
65,536	4.47 GiB	1.83 GiB	2.44×
131,072	8.47 GiB	3.66 GiB	2.32×